ABSTRACT. In this work we present a new approach for simulating and estimating temperatures in urban environments solely from very high resolution satellite stereo imagery. For this first a dense digital surface model (DSM) is calculated from two or more satellite stereo images. Deriving a digital terrain model (DTM) from this DSM allows for extraction of man-made high objects and also for plants and trees. Using also the spectral information from the satellite imagery allows for classification of the scene to different surface materials. The main classes derived for a urban scene contain water, tree, grass, soil, road, tiles and concrete. For each of the classes physical constants like the heat-capacity c in [kJ/kg/K] , the heat-transport λ in [W/m/K], the density of the material ϱ in [kg/dm3 ], the absorption ϵ (in %) and diffusion resistance rl in [s/m] are derived and used for the further simulation. Also for preparing the simulation the sky-view of each pixel in the resulting DSM is calculated together with shadows for each 300 seconds for a whole day. Afterwards the simulation runs several times for the day until stability is reached. The results are temperature maps for each 5 minutes for a whole day of 24 hours. Finally the results are compared with independent temperature measurements from satellite.

ABSTRACT. Laser scanning and photogrammetry are among the most common techniques for accurately surveying and documenting architectural and cultural monuments. Integrating these two methods makes it possible to produce high-resolution 3D representations, which facilitate detailed documentation and analysis of historic buildings and structures. This paper proposes a method for extracting feature points from a point cloud obtained through laser scanning. These feature points can be used as control points in photogrammetric processing, particularly for aligning images captured by aerial means. This approach allows for leveraging the precision of laser scanning in the photogrammetric processing of images for structures where it is challenging to create and measure artificial control points.

ABSTRACT. Accessibility to essential services in rural areas is critical for quality of life but is often constrained by population dispersion, infrastructural limitations, low availability and demographic shifts. This study examines the interplay between urban development, demographic change, and accessibility to amenities in the rural Westhoek region of Belgium between 2015 and 2019. Using high-resolution geospatial and population data, we analyze changes in built-up areas, population distribution, and accessibility to essential services. The findings reveal that while urbanization and population density correlate with accessibility, improvements of walkability often lag behind development. Regression models highlight a strong relationship between accessibility and degree of urbanization. However, change analyses show weak correlations, indicating that urban growth does not consistently lead to immediate accessibility gains. This lag underscores the need for synchronized planning of rural expansion and service development to ensure equitable access for all rural residents.

ABSTRACT. Blind Spectral Unmixing (SU) in Hyperspectral Image (HSI) analysis benefits from integrating spatial information to improve endmember and abundance estimation. Gabor Filtering (GF) effectively extracts spatial features, but combining these with spectral bands is challenging. We propose a method that combines spectral and Gabor spatial features to enhance SU performance. First, we use the Laplace Score (LS) method to select important spectral bands. Gabor filters are applied to these bands for texture extraction. A multi-view Convolutional Neural Networks (CNNs) merges the features in a shared latent space, and an Autoencoder (AE) performs SU. Experiments on two real HSI datasets demonstrate the method's superiority over existing approaches.

ABSTRACT. Floods are among the most devastating natural disasters, causing severe human and economic losses. Their occurrence frequency has been increasing progressively. Hence, effective and reliable flood analysis is essential for mitigating catastrophic losses. In this regard, Synthetic Aperture Radar (SAR) satellites are invaluable tools for providing large-scale images aimed at flood mapping under all weather conditions. Methods specifically designed for SAR image-based flood mapping are commonly developed under the assumption that the training (source) and test (target) data are sampled from the same distribution. However, in many real-world scenarios, these distributions often differ due to factors such as geographic location and incident angle, leading to distribution shifts (a.k.a. domain shift), which ultimately degrades model performance. In this study, we investigate domain generalization approaches in combination with segmentation networks for the purpose of SAR based flood mapping. During the model’s training, each flood event is treated as a distinct source domain, with the objective of minimizing the domain shift among them to obtain a more robust model for unseen flooding events. Experiments conducted on the Sen1Floods11 dataset demonstrates an improvement in segmentation performance, with domain generalization approaches 3% in terms of IoU and F1 scores.

ABSTRACT. Urbanization, characterized by physical growth and infrastructure expansion, remains a global phenomenon with profound economic, environmental, and social implications. Traditional methods of monitoring construction activity—key indicators of urban and economic development—often suffer from delays and inconsistencies in reporting. This study introduces a Satellite-based Building Activity Indicator (SBAI) leveraging Sentinel-2 imagery and machine learning to provide highly resolved, spatiotemporal data on urban construction activities. The SBAI identifies new urban developments with high accuracy and highlights the ability to track seasonal and project-specific construction trends. Comparison with official statistics on construction activity validates its reliability, while its granularity offers enhanced insights into intra-year variations and localized urban growth dynamics. The SBAI demonstrates significant potential as a tool for national statistical frameworks, offering timely and detailed data to support responsive and informed decision-making at local and national levels.

ABSTRACT. The urban exposome—comprising environmental exposures like air pollution, noise, and green space access—significantly impacts cardiovascular diseases (CVD). While remote sensing has emerged as a critical tool for mapping and analyzing urban exposures, its potential remains underutilized within neighborhood-level analyses (e.g. at a 100 m resolution). This paper explores the intersection of the urban exposome and CVD, using remote sensing to bridge knowledge gaps. By highlighting CVD as an exemplary group of diseases, we outline existing global evidence on urban environmental exposures and CVD risk factors, emphasizing gaps in understanding specific pathways. Our focus turns to Germany, where urban infrastructure and health contexts offer unique opportunities and challenges for exposome research. Through a review of German population-based cohort studies, we assess their incorporation of spatial data and identify limitations in linking urban environmental exposures to CVD risk factors. This analysis underscores the need for integrating high-resolution remote sensing data into localized studies, paving the way for tailored public health interventions and urban planning strategies.

ABSTRACT. This study addresses the need for accurate tree inventory monitoring in urban areas to support planning and environmental efforts, by leveraging advances in deep learning for single tree detection. Using very high resolution aerial images of a residential area in Munich, we compared the performance of two object detection algorithms, YOLOv8 and DeepForest, on the urban tree detection task. The ground truth data consisted of hand drawn tree bounding boxes, and the performance was monitored mostly by F1 score measures. Subsequently, efforts concentrated on improving DeepForest’s performance through fine-tuning. The enhancements yielded a notable increase in the model’s F1 score from 0.7365 to 0.8030, indicating the effectiveness of these techniques. These findings further underline the potential of deep learning for urban tree detection and highlight the viability of models like DeepForest for creating accurate urban tree inventories, with promising avenues for future enhancement.

ABSTRACT. Smart Cities are those that use Information and Communication Technologies (ICT) to help make cities work better. They are existing cities in developed countries that are in the process of intensive development to improve the quality of life of their citizens, environmental sustainability, the effectiveness of their public services and the overall efficiency of the management of their urban areas. The emergence and multiplication of cities classified as smart in developed countries following the adoption of this new approach to urban development is a reflection of this new concept. The aim is to improve the lifestyle and quality of life of city dwellers through the use of new technologies on the basis of an ecosystem, with the integration of both objects and services. The aim is to make each Smart City more adaptable and efficient, and to improve the lifestyle and quality of life of its inhabitants. In this regard, the MIMAC method has proved to be the most appropriate. It is aimed at improving the strategy and proposing an appropriate Action Plan to re-organise the City of Sfax into a future Smart City. This has led to its use in the continuous process improvement model underlying the digital revolution.

ABSTRACT. Urban remote sensing is a pivotal tool for monitoring and achieving Sustainable Development Goals (SDGs), especially in urban areas facing rapid growth and environmental challenges. This study leverages urban remote sensing to assess and analyze sustainable urban development in the Gabes region of Tunisia, focusing on its societal impacts. Employing high-resolution satellite data, we examine key indicators related to SDG 11 (Sustainable Cities and Communities), particularly in urban expansion, green space coverage, and infrastructure quality. Results show significant disparities within the region, revealing that only 35% of Gabes’s urban population has convenient access to public green spaces. Furthermore, urban expansion over the past decade has led to a 12% loss in agricultural land, impacting both ecological balance and local livelihoods. This study underscores the importance of integrating remote sensing in urban planning processes to achieve SDG targets and mitigate adverse societal impacts.

ABSTRACT. The MUSE (Measuring U-Space Social and Environmental Impact) project aims to develop an innovative framework and toolset for measuring and forecasting environmental and social Key Performance Indicators (KPIs) related to Urban Air Mobility (UAM), with a focus on assessing drone trajectory impacts, noise emissions, visual pollution, and population exposure in urban areas. Within the scope of the MUSE project, this study addresses the challenge of accurately detecting pedestrians in complex urban environments using high-resolution satellite imagery with the goal to later evaluate the visual pollution of drones in Urban area. The pilot city is Madrid and we used the highest resolution of optical satellite available on the market to resolve the challenge.

ABSTRACT. Our knowledge on long-term urbanization trends prior to the 1970s is sparse, mainly due to a lack of harmonized, digital geospatial data. However, such knowledge is crucial to better understand settlement dynamics, and to provide more reliable, quantitative data sources for long-term population modeling. Herein, we outline a data integration effort that leverages multi-modal geospatial data to provide reliable estimates of built-up surface and its spatial distribution back to the 1950s, in a scalable way, making use of recent Earth observation data, but also historical maps, aerial imagery, administrative data and other geospatial data sources. First experiments show promising results in both, the data-abundant global North as well as for study areas in the global South.

ABSTRACT. The settlement landscape forms an urban-rural continuum between high-density urban centers and open rural areas. Translating semantic settlement types such as urban core or suburban area into space, however, is challenging, as there are no generally accepted definitions for this. In this work, we use expert knowledge, remote sensing, open geodata and machine learning to derive settlement types for Germany. We combine a set of reference points for six settlement types collected by experts with a data cube of 699 descriptive features from remote sensing and other data sources to create a supervised random forest classification model. Our results show a continuous map of settlement types in Germany with high accuracy and high spatial detail. The largest uncertainties in the classification model arise in the distinction between suburban, periurban and rural areas, where the manual classification by experts was also not unambiguous. The result shows that 6.2 % of Germany is covered by urban or suburban settlement types, which are home to 66.8 % of the population.

ABSTRACT. Urbanization is a human-made process and has a direct impact on regional climate. In this study, we assess the effect of urbanization on land surface temperature using remote sensing images and over Oran city in Algeria. Satellite data from Landsat sensors were used to spatially quantify land surface temperature (LST) using an improved split-window algorithm (SWA), and quantify surface heat island intensity (SHI) using the average surface temperature of each district and neighborhood within the city compared to the rural surface temperature outside the city. The quantitative assessment is based on surface temperature maps retrieved for 5 years, from the year 2017 to 2021. Results showed that the LST was high over the slum built-up areas, while low over the individual and old built-up areas. With the same way, SHI is largest for neighborhoods with scarce vegetation that have a high fraction of impervious surface.

ABSTRACT. Remote sensing plays a critical role in assessing urban landscapes. Despite advances in spatial and temporal resolution, satellite-based remote sensing alone cannot capture fine-scale variations in aspects such as temperature, humidity, air quality and others. This shows a need for in-situ measurements to accompany satellite data, which cannot always be satisfied by the limited network of existing fixed measuring stations. This paper introduces a cargo bicycle outfitted with environmental sensors to address this gap, offering high spatial and temporal resolution for in-situ measurements. We detail the design and implementation of the cargo bike with its sensors as well as the data processing systems developed for this purpose. The results of a measurement campaign conducted in June 2024 demonstrate the system's effectiveness. This approach enables a more comprehensive understanding of urban environmental dynamics, paving the way for enhanced environmental monitoring in cities.

ABSTRACT. Climate change is a critical global challenge that demands innovative strategies for effective mitigation. This study performs a bibliometric analysis of artificial intelligence applications in climate change research, utilizing VOSviewer software to examine 152 publications from the Scopus database. The analysis indicates a significant rise in research activity, with prevalent keywords such as "climate change," "artificial intelligence," and "machine learning" highlighting the key focus areas. The findings emphasize the vital role of AI technologies in enhancing sustainability and addressing climate challenges, advocating for solutions that integrate environmental, social, and economic goals. Notable sources, including "machine learning and artificial intelligence," "AI for climate change and environmental sustainability," and "environmental chemistry letters," have been instrumental in shaping the field. Geographically, influential contributions have emerged from India, the United States, and the United Kingdom, with the University of Tabriz and Menoufia University identified as leading institutions. Key authors such as Nourani, V. (four publications), along with highly cited researcher Huntingford, C. (204 citations), significantly advance the literature. These findings not only illuminate future research directions but also inform the development of effective AI-driven solutions for sustainable climate practices. The study highlights the crucial integration of AI in climate action, promoting collaborative approaches to enhance resilience and sustainability in the face of environmental challenges.

ABSTRACT. An accurate and timely characterization of built-up areas is essential in making cities and human settlements safe, resilient, and sustainable. Dual-polarized (dual-pol) Synthetic Aperture Radar (SAR) data, commonly used for urban analysis, primarily relies on backscatter intensity, which lacks adequate built-up target characteristics in complex urban scenarios. This study utilizes a set of dual-pol descriptors specifically designed to discriminate built-up areas from other land cover targets, facilitating enhanced characterization of diverse built-up targets in complex urban environments. The applicability of these descriptors to Sentinel-1 Ground Range Detected (GRD) SAR data accessible on the Google Earth Engine (GEE) platform enables efficient global built-up area characterization with frequent updates.

ABSTRACT. In this paper, we present a novel method for urban area extraction using Sentinel-1 synthetic aperture radar (SAR) imagery. The method utilizes an unsupervised K-means classifier applied to the total backscatter power from dual-polarimetric channels in Sentinel-1 SAR data. To validate the approach, we tested it on two datasets: one from the Algiers region and another from the San Francisco region. Results demonstrate that the proposed method effectively distinguishes urban areas, showing strong potential for accurate urban mapping. Additionally, it reliably identifies the three principal scattering mechanisms on the Earth’s surface (double-bounce, volume, and surface scattering) offering a comprehensive tool for urban and environmental applications.

ABSTRACT. The accurate localization of multiple scatterers within a single-resolution cell constitutes a challenging task in the context of Tomographic Synthetic Aperture Radar (TomoSAR). In order to produce comprehensive three-dimensional maps, Generalized Likelihood Ratio Test (GLRT)-based detectors can be used. The most precise 3D point clouds can be achieved using the Support GLRT variant, at the expense of computationally intensive multidimensional grid search for finding the scatterers’ positions. In this paper, we propose Gradient Descent GD-GLRT, a gridless approach for effective yet precise scatterers’ detection in TomoSAR. The proposed detector uses the Fast Sup-GLRT for initializing the search followed by local descent optimization that operates in the continuous domain. The experimental results on both synthetic and real data show the effectiveness of the GD-GLRT in resolving the off-grid issue while decreasing computational costs.

ABSTRACT. Building segmentation in urban areas is essential in fields such as urban planning, disaster response, and population mapping. Yet accurately segmenting buildings in dense urban regions presents challenges due to the large size and high resolution of satellite images. This study investigates the use of a Quanvolutional pre-processing to enhance the capability of the Attention U-Net model in the building segmentation. Specifically, this paper focuses on the urban landscape of Tunis, utilizing Sentinel-1 Synthetic Aperture Radar (SAR) imagery. In this work, Quanvolution was used to extract more informative feature maps that capture essential structural details in radar imagery, proving beneficial for accurate building segmentation. Preliminary results indicate that proposed methodology achieves comparable test accuracy to the standard Attention U-Net model while significantly reducing network parameters. This result aligns with findings from previous works, confirming that Quanvolution not only maintains model accuracy but also increases computational efficiency. These promising outcomes highlight the potential of quantum-assisted Deep Learning frameworks for large-scale building segmentation in urban environments.

ABSTRACT. Building height estimation is a challenging and essential task in various fields such as disaster risk management, urban planning, and change detection evaluation. However accurate measurements of building heights are challenging due to the different characterization and complexity of urban structures in cities. One key issue to be addressed is whether building height estimation is more effectively performed by considering each pixel as an independent unit or by analyzing the building as an integrated object composed of multiple pixels. This distinction is crucial, as it can substantially impact both the accuracy of results and the practical applications of the analysis. In this paper, we present two deep learning-based methodologies for estimating building heights using a single high-resolution COSMO-SkyMed image. The first methodology employs an Attention-UNet model and functions as a pixel-wise approach, while the second utilizes ResNet101 as its core architecture and operates in an object-based manner. The urban area of Milan, Italy, was selected as the study area for this research. The results indicate that the first methodology achieves a lower Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) compared to the second methodology. The comparative results of these two methodologies are substantial and offer valuable insights for decision-makers, providing a clearer understanding of urban environments. The code used in this study can be made publicly available on GitHub following potential acceptance of the work.

ABSTRACT. Unmixing land cover optical properties (OP) from coarse spatial resolution remote sensing images is a critical methodological challenge for microclimate and energy balance research. We propose the novel Unmixing Spectral method (US-DART) using 3D differentiable radiative transfer model DART to extract endmember OPs from shortwave (SW) mono-/multi-spectral RS images. It has 4 modules: (1) pure pixel selection, (2) linear spectral mixture analysis, (3) gradient iterative refinement, and (4) spectral correlation. It has 3 inputs: a reflectance image, a spatially-classified 3D scene model, and relevant DART parameters (e.g., spatial resolution, skylight ratio, ...). It generates an OP map per type of scene element. It is validated for 3D vegetation and urban scenes quantitatively with DART-simulated satellite image (median pixel reflectance relative error: about 0.1%) and qualitatively with Sentinel-2, PlanetScope, and Google Map Images. Performance is better for opaque surfaces (about 1%) such as roofs than for translucent materials (1-5%), such as leaves. By better extracting OPs from coarse resolution SW imagery US-DART substantially advances remote sensing applications, such as simulations of time series of SW albedo and radiation balance.

ABSTRACT. Airborne lidar point clouds allow a dense, accurate 3D acquisition of landscapes at large scales. With the increasing number of sensors and applications, multiple surveys can now cover the same area, opening the field for effective and realistic 3D change detection. We propose here an operational approach to capture and quantify various types of changes under several acquisition configurations. A fuzzy logic approach leverages both the geometric and semantic change information that was preliminarily captured, subsequently allowing to infer the types of change at a 2m voxel level. Experiments show the validity of our solution, without supervision, and heuristic parameter tuning.

ABSTRACT. In this paper, we present a parallel computing strategy for the detection of single scatterers in SAR tomography, specifically targeted to multi-node, multi-GPU platforms. In particular, we refer to an existing processing scheme based on the Generalized Likelihood Ratio Test (GLRT), which is considered a reference canonical problem for our investigation. To tackle this problem, a dedicated parallel algorithm is developed according to High-Performance Computing (HPC) methodologies. To quantitatively demonstrate the benefits of the multi-level parallelism incorporated into the proposed algorithm, an experimental analysis using real SAR data is conducted on a large distributed computational platform. The results show significant speedup, indicating that the proposed solution maintains excellent performance as both the problem size and computational resources increase.

ABSTRACT. In this study, we explored the performance impact of incorporating country-level text-based geographical information into a large-scale vision language model, fine-tuned for the captioning of optical remote sensing images. We hypothesized that a model trained with country-level textual geographical context along with visual scenes would enhance its captioning capabilities when confronted with images from previously unseen countries or even continents, coupled with their respective geographical context. A large language and vision assistant (LLaVA) was fine- tuned using optical images from European countries and tested on images from other continents to evaluate its generalization capabilities. Here we report results of experiments conducted across 175 countries via the newly published Skyscript dataset, demonstrating that even superficial geographical information obtained from Wikipedia articles can mitigate the cross-country domain shift by several points in terms of accuracy score. This multimodal approach, combining textual geographical context with visual data, shows significant potential for improving the generalization capabilities of vision language models in tasks involving diverse and previously unseen geographical regions.